Sunteți pe pagina 1din 10

Available online at www.sciencedirect.

com

ScienceDirect
Materials Today: Proceedings 4 (2017) 19471956 www.materialstoday.com/proceedings

5th International Conference of Materials Processing and Characterization (ICMPC 2016)

RSM and Fuzzy logic approaches for predicting the surface


roughness during EDM of Al-SiCp MMC
R. K. Bhuyan*, Shalini Mohanty, B.C. Routara
School of Mechanical Engineering, KIIT University,Bhubaneswar, Odisha

Abstract

Surface roughness is the one of the critical performance parameter that has been effect on several mechanical properties of
machined parts like friction, wear, light reflection, heat transmission, lubrication, electrical conductivity, etc. Hence this paper
present the surface roughness like Ra,Rq and Rz of Al-SiCp metal matrix composite (MMC) during electric discharge machining
(EDM). In order to achieved the desired surface roughness the experiment has been planed based on central composite design
(CCD) method with three EDM parameters such as pulse-on time (TON), peak current (Ip), and flushing pressure (Fp). In this
paper describes the mathematical modeling for response surface methodology (RSM) and fuzzy logic modeling technique to
prediction themeasuring surface roughness (Ra,Rz and Rq) of Al-SiCpmetal matrix composite (MMC).Also the performance of the
experimental result is compared with the developed fuzzy models and RSM mathematical models and it clearly indicates that the
Fuzzy models provide more accurate prediction in compared to the RSM models. Finally the Analysis of Variance (ANOVA)
technique is carried out to check the significance of the models and study the effect of process parameters.
2017 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of Conference Committee Members of 5th International Conference of Materials
Processing and Characterization (ICMPC 2016).

Keywords:Aluminium metal matrix composites;RSM; Fuzzy logic; ANOVA

1.Introduction

Surface finish is an important attribute of any machining operation. To achieve this high quality, surface finish,
dimensional accuracy work piece, high production rate, economy of machining in terms of cost saving and increase
the performance of the product with reduced environmental impact are the main and effective challenges of modern
Non-traditional machining process. [1-2]. EDM machining consider as a Non-traditional machining processes is the
most widely and successfully applied for high precision surface finishing process of various conductive materials
regardless of their mechanical properties .Basically In EDM the surface roughness mainly depends on the
bombardment of high energy electron on the electrode surface as result near about a temperature 10,0000C is
developed between the tool and work piece during the sparking. At this high temperature the material are melts and
vaporized with leaving the crater on the surface. Also due to flushing of dielectric the surface near the electrodes
rapidly cool that sparking zone . This rapid cooling and heating cause distinctive surface morphology on the
machined surface. [3-4].
*
Corresponding author. Rajesh Kumar Bhuyan, Tel.: +91-9938513183.
E-mail address:rajesh_bhuyan001@rediffmail.com

2214-78532017 Elsevier Ltd. All rights reserved.


Selection and peer-review under responsibility of Conference Committee Members of 5th International Conference of Materials Processing and
Characterization (ICMPC 2016).
1948 R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956

1.1. Literature survey

Meanwhile to improve surface roughness in EDM Process, it is necessary to select the appropriate process
parameters to get the highest performance for desired dimensional accuracy with improved the surface quality.
Among the several attempts by the researcher Pargunde et al. [5] developed an Al-based SiC MMC using stir
casting method. They fabricated the composite with Aluminum alloy 98.41% pure and SiC (360-grit) with a varying
weghit of 5%, 10%, 15%, 20%, 25%, 30% respectively. They observed that the maximum Hardness, Density
(gm/cc) and impact strength at 25% SiC weight fraction. Satyanarayen et al. [6] used stir casting technique to
prepare the aluminum LM25-SiC composites with added the 3%, 6% and 9% of weight percentage of SiC. They
found that tensile strength and hardness increase with percentage SiC increase. Puhan et al. [7] investigate the mach-
inability effect of Al-SiC MMC prepare by powder metallurgy process during EDM. They are used four process
parameter like discharge Current, pulse duration, duty cycle, and flushing pressure on the material removal rate, tool
wear rate, surface roughness (Ra) and circularity. They optimize the process parameters by principal component
analysis (PCA) with fuzzy inference system. Also they stated that the hardness of MMC is increasing with
increasing weight % of SiC in the composite. Pradhan et al. [8] planned central composite design (CCD) method to
explained the effect of four process parameter like peak current, pulse on time, pulse off time and discharge voltage
on the surface roughness (Ra) during EDM of AISI D2 steel. They predicted the Ravalue by using the mathematical
model equation by RSM and the performances of the parameters are analyzed by ANOVA.

Rajesha et al.[9] using Taguchi L9 technique optimize the input process parameter like gap current, pulse on-time
and pulse off-time on the output surface roughness (Ra and Rz) of inside the hole and near the hole while EDM
drilling of Al 7075 MMC. Vishwakarma et al.[10] using central composite design method to optimized the process
parameter like pulse on time, gap voltage, flushing pressure, input current and duty cycle on the material removal,
surface roughness during EDM of AISI 4140 grade steel alloy. They developed the mathematical model by using
RSM to predicate the experimental result of the output. Khalidet al.[11] optimized the process parameter like
current, pulse on time and pulse off time on the output material removal rate, tool wear rate and surface roughness
(Ra) using fuzzy logic evolutionary strategies technique during EDM of the three different materials such as
stainless steel, C40 Carbon steel and SKD61 and stateproposed that this techniques is a benchmark to solve the
multi-objective problems. Kohli et al.[12] compare the experimental results of material removal rate with the result
generated by fuzzy model during die sinking EDM process of Medium Carbon Steel (AISI 1040) as work-piece.
They described that proposed fuzzy model result is close agreement with the experimental results.

Rao et al.[13]investigate the effect of current , open-circuit voltage, servo duty cycle on the responses like
material removal rate,tool wear rate ,surface roughness (Ra) and hardness HRB during EDM of AISI 64430
aluminum alloy. They predicted and compare the experimental result by using Fuzzy logic approach and stated that
the variation predicted values can be reduced by considering more fuzzy member ship functions with the selected
range of parameters. Shandilya et al.[14] using RSM and artificial neural network technique to compare the average
cutting speed of Al 6061-SiCp MMC during wire electric discharge machining. They observed that the ANN models
provide more accurate prediction in compared to the RSM models. Rodic et al.[15] applied Fuzzy logic and neural
net work modelling to compare the experiment result and predicated result of the surface roughness (Ra) during
EDM of manganese alloyed cold-work tool steel. They stated that neural net work technique predicated more
improvement forecast than fuzzy logic.

Based on the literature review this paper is focused on the EDM process with three input variables such as peak
current, pulse on time and flushing pressure of the material Al -SiC 12% MMC. The output responses are the three
surface roughness factors such as Ra,Rq and Rz. The objective of this paper is to compare the experimental result
with the predicted result by RSM and Fuzzy logic technology.
R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956 1949

2. EXPERIMENTAL DETAILS
2.1 Work-piece materials

The metal matrix composite materials (MMC) are prepared by stir casting to fabricate the Al-SiCP MMC. The
commercial pure aluminum with 12% weight fraction of silicon carbide particle (SiCp) with the average particle
grain size is 0.0228 mm are used for the made up of the Al-12%SiCP MMC. The original size of the sample
specimen is prepared approximately 40 mm in diameter and length approximately 360 mm. After that each work
piece are prepared around 40 mm in diameter and 10 mm in thickness.

2.2 Design of experiment

The experiment is conducted by Central Composite Design (CCD) methodto run the 20 no of experiments. To
run the experiments three process parameters are consider i.e Peak current (Ip) in Ampere (A), Pulse on time (Ton) in
Microsecond (s) and Flushing pressure (Fp) in Kg/cm2. Based on the literature survey the CCD method the ranges
of the parameters are selected with different levels as shown in the Table 1.

Table 1. Process parameter and their levels

Levels
Parameters
-1.682 -1 0 1 1.682
Peak current (Ip) 3.2 10 20 30 36.8
Pulse on time (Ton) 116 150 200 250 284
Flushing pressure (Fp) 0.164 0.3 0.5 0.7 0.836

2.3 Experimental method & Results

The experiments are carried out by Electrical Discharge machining of model MIC-432CS CNC
manufactured by ECOWIN at CIPET, Patia, Bhubaneswar. The tools are made of electrolyte copper electrode with
average diameter 25.4 mm for each experiment. The depth of machining of each work piece is 2mm .The surface
roughness of the machining surface Ra, Rq, and Rz are measured by MITUTOYO Surface roughness tester. The Ra
is measured the arithmetic average roughness of the evaluation length of the machining surface. It measured by the
following relation [16-17].
L


Ra = Y ( X )dx
0
where L=evaluation length
Y=The ordinate of the profile curve
Rq known as Root-mean-square roughness.
It measured as the arithmetic mean of the departure of the roughness profile from the mean line.

Y (X ) dx
2
Rq =
0

Rz average distance between the highest peak and lowest valley in each sampling length
1 s
Rz = Rti
s i =1
where S is the number of sampling lengths, and Rtiis the maximum height of the ith sampling length.
1950 R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956

Fig. 1 Surface roughness profile

The figure 1 show Surface roughness profile of the selected responses. Also the experimental result of responses is
shown in Table 2.

Table 2. Experimental results using DOE


Pulse on
Test .no Peak Current(A) Flushing pressure(Kg/cm2) Ra Rq Rz
time(S)

1 150 10.0 0.300 10.527 10.964 50.154


2 250 10.0 0.300 8.993 11.785 55.774
3 150 30.0 0.300 15.980 22.951 98.225
4 250 30.0 0.300 16.979 29.299 124.071
5 150 10.0 0.700 10.154 12.912 64.950
6 250 10.0 0.700 14.121 15.451 62.581
7 150 30.0 0.700 16.485 20.298 91.720
8 250 30.0 0.700 26.813 33.972 134.240
9 116 20.0 0.500 11.716 17.359 84.865
10 284 20.0 0.500 19.748 24.751 107.915
11 200 3.2 0.500 8.338 8.540 44.189
12 200 36.8 0.500 24.703 30.613 126.631
13 200 20.0 0.164 12.066 25.858 110.607
14 200 20.0 0.836 20.357 21.168 94.008
15 200 20.0 0.500 13.448 20.221 104.162
16 200 20.0 0.500 13.148 20.121 104.031
17 200 20.0 0.500 13.498 21.021 103.990
18 200 20.0 0.500 13.546 19.991 104.317
19 200 20.0 0.500 13.126 20.311 104.069
20 200 20.0 0.500 13.997 20.145 104.751

3. Methodology

3.1 Response surface Methodology

Response surface methodology (RSM) is a set of mathematical and statistical techniques used to build the
empirical quadratic model equation.The primary objective of the RSM is to make a relationship between the input
process parameter and the outputs responses. The relationship between input process parameter and the outputs
responses arecommonly represented by a function ( f ) i.e.
R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956 1951

R = (x1 , x2 , x3 ........................xn ) (1)

Where R is defined as the response and x 1 , x 2 , x 3 .......... .......... ....x n are the input process parameters.
Based on the above equation the second order polynomial regression model can be represented in the form of
quadratic equation for the three input process parameter such as
R = 0 + 1 x1 + 2 x2 + 3 x3 + 4 x1 2 + 5 x2 2 + 6 x3 2 + 7 x1 x2 + 8 x1 x3 + 9 x2 x3 (2)

Where 0 is a constant and 1 , 2 , 3 are the coefficient of linear effect of the input process parameters like
x1 , x2 , x3 respectively. For 4 , 5 , 6 are the coefficient of the square effect of input parameter themselves such

as x , x , x32
2
1
2
2 and 7 , 8 , 9 are the coefficient of liner by liner interaction among the input parameter like x1 by
x2 , x1 by x3 and x2 by x3 respectively .The coefficients value are get by an appropriate method such as the least
square method or Minitab Software.[18-19]

Based on the above discussion the following equations are represented the surface roughness of the responses and
the table 3 show the predicted result by the RSM.
Ra = 38.3120 - 0.1813Ton - 0.5019 Ip - 50.8117 Fp + 0.0002Ton 2 + 0.0081Ip 2 (3)
+ 17.3918 Fp 2 + 0.0022Ton * Ip + 0.1854Ton * Fp + 0.3490 Ip * Fp

Rq = 22.7281 - 0.0709401Ton + 0.207636 Ip - 37.4172 Fp - 0.0000Ton 2 - 0.00626 Ip 2 (4)


2
+ 19.2077 Fp + 0.004165Ton * Ip + 0.113054Ton * Fp - 0.224563Ip * Fp

Rz = -39.0692 + 0.0526928Ton + 3.31285 Ip + 63.7562 Fp - 0.00186262Ton 2 (5)


2 2
- 0.0854687 Ip + 64.0015 Fp + 0.0162788Ton * Ip + 0.108550Ton * Fp - 1.12121Ip * Fp

3.2.Fuzzy Logic Modeling for Electrical Discharge Machining Process

Fuzzy logic is a multi-reasoning logical concept where the evaluation based on true/false, yes/no,and
high/low etc In the fuzzy-logic modeling the reasoning are defined in terms of human linguistic that is the common
reasoning for the decision-making of the situation problems. The linguistic variables are defined as in terms of
extremely small, very small, small, medium, less high, high, very high, very very high and extremely high etc.In the
fuzzy Logic Mamdanimodeling system mainly consists of data definition, fuzzification, implication, aggregation
and defuzzification. The data definition defined all the inputs and outputs data are to be consider for the selected
model. The fuzzification and implication the users developed the membership functions in terms of human linguistic
variables to generate the fuzzy rules. The fuzzy rules are described in the Mamdani model by if-then rule .Next in
the aggregations process mixed up the all output rules and the defuzzification process converts the aggregation
output to a single predicted value In this study three input parameters such as Pulse on Time (Ton), Peak current (Ip)
and flushing pressure (Fp) and three output responses likely Surface roughness (Ra,Rq,Rz) are selected for this fuzzy
model. The modelling is carried out by fuzzy interface system (FIS) as shown in Figure 2.The input and output
parameters are represented by linguistic variables or fuzzy membership function in the triangular shape. For the each
inputs Ton, Ip and Fp threelinguistic value or membership likely minimum, medium and maximum and for the output
nine linguistic variables such as extremely small (ES),very small (VS), small (S), medium (M), less high (LH), high
(H), very high (VH), very very high (VVH) and extremely high (EH). [13][20-21]
1952 R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956

Fig. 2 .Fuzzy logic model of surface roughness


4. Results and discussions

4.1. Analysis of Variance (ANOVA)

The result of ANOVA test of the surface roughness is shown in Table 3-5. From the table it is concluded
that if the P value is less than 0.05 then the linear, square, interaction of the process parameter are significant and
95% confidences level. The R2 and Adj R2 are indicating the measure result goodness of fit for the model. It
significant that if it closes to unity the experimental result is better and fit for the model. The value of R2 and Adj R2
are for for Ra 97.19%, 94.66% for Rq 95.75%, 91.93% and Rz 95.38%, 91.21% respectively. The R2 and Adj R2
results concluded that the develop model is statistically considerable [22-25].

Table 3.Analysis of Variance for Surface Roughness (Ra)


Source DF SeqSS AdjSS Adj MS F P Remark
Linear 3 379.611 30.998 10.3326 8.16 0.005 Significant
Square 3 17.067 17.067 5.6889 4.49 0.030 Significant
Interaction 3 41.277 41.277 13.7589 10.87 0.002 Significant
Residual Error 10 12.659 12.659 1.2659
Total 19 450.613

Table 4.Analysis of Variance for Surface Roughness (Rq)


Source DF SeqSS AdjSS Adj MS F P Remark
Linear 3 720.899 16.600 5.5332 1.59 0.252 Not Significant
Square 3 15.825 15.825 5.2749 1.52 0.269 Not Significant
Interaction 3 46.543 46.543 15.5143 4.47 0.031 Significant
Residual Error 10 34.742 34.742 3.4742
Total 19 818.009

Table 5.Analysis of Variance for Surface Roughness (Rz)


Source DF SeqSS AdjSS Adj MS F P Remark
Linear 3 10040.6 326.4 108.79 1.88 0.196 Not Significant
Square 3 1288.4 1288.4 429.47 7.44 0.007 Significant
Interaction 3 579.7 579.7 193.22 3.35 0.064 Not Significant
Residual Error 10 577.4 577.4 57.74
Total 19 12486.1
4.2 Comparison of the RSM and Fuzzy logic models

To compared the RSM and Fuzzy logic method the predictive result and the experimental results are
analyzed by their percentage of absolute error of the responses. The equation 6 is used for to find the percentage of
error on each experimental result. The Table 6-8 and the figure 3showthe average absolute percentage of error of
each response. It is analyzed that experiment 2 has been less Ra,Rq and Rz value .Hence the corresponding process
R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956 1953

parameter for the experiment no.2 is the bestinput process parameter i.e. Ton =250 s, Ip=10A and Fp=0.300 Kg/cm2
respectively.The average absolute percentage of error for Ra by RSM is 0.163 and by Fuzzy logic is 0.043 but for
Rqby RSM is 0.355 and for fuzzy logic is 0.184 although for Rz by RSM is 0.395 and by Fuzzy logic is 0.293
respectively.
(Predicted result - Experimental result)
% Absoulte Error = 100
Predicted result

(6)

Fig. 3. Comparison average error of the surface roughness

Table 6. RSM and Fuzzy logic model for predictions the experimental result of Ra

Test .no Experimental result RSM prediction Fuzzy logic prediction


1 10.527 10.6762 10.577
2 8.993 8.7400 8.829
3 15.980 15.8447 15.425
4 16.979 18.3555 16.970
5 10.154 9.8268 10.577
6 14.121 15.3055 15.577
7 16.485 17.7873 17.422
8 26.813 27.7130 26.971
9 11.716 11.6329 11.316
10 19.748 18.3441 18.894
11 8.338 8.3951 8.569
12 24.703 23.1589 24.945
13 12.066 11.8946 11.936
14 20.357 19.0414 12.492
15 13.448 13.5045 13.207
16 13.148 13.5045 13.207
17 13.498 13.5045 13.207
18 13.546 13.5045 13.207
19 13.126 13.5045 13.207
20 13.997 13.5045 13.207
Average % of error 0.163 0.043
1954 R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956

Table 7. RSM and Fuzzy logic model for predictions the experimental result of Rq

Test .no Experimental result RSM prediction Fuzzy logic prediction

1 10.964 13.779 10.9839

2 11.785 12.600 11.9839

3 22.951 24.069 23.8724

4 29.299 31.221 30.6154

5 12.912 12.380 13.6656

6 15.451 15.723 15.6656

7 20.298 20.873 20.6656

8 33.972 32.548 33.8583

9 17.359 15.662 17.6097

10 24.751 24.478 24.4781

11 8.540 7.204 8.8725

12 30.613 29.979 30.0681

13 25.858 22.559 24.6142

14 21.168 22.498 20.875

15 20.221 20.360 20.0781

16 20.121 20.360 20.0781

17 21.021 20.360 21.0781

18 19.991 20.360 19.0781

19 20.311 20.360 20.0781

20 20.145 20.360 19.0781

Average % of error 0.355 0.184


R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956 1955

Table 8. RSM and Fuzzy logic model for predictions the experimental result of Rz

Test .no Experimental result RSM prediction Fuzzy logic prediction

1 50.154 61.949 56.4353

2 55.774 59.672 56.6134

3 98.225 101.940 99.1327

4 124.071 132.221 123.3289

5 64.950 63.879 65.9489

6 62.581 65.944 65.9489

7 91.720 94.900 93.9489

8 134.240 129.523 133.1166

9 84.865 77.789 81.8959

10 107.915 104.960 107.2735

11 44.189 36.895 37.5652

12 126.631 123.894 125.5116

13 110.607 97.614 108.9145

14 94.008 96.969 95.3392

15 104.162 104.517 104.5652

16 104.031 104.517 104.5652

17 103.990 104.517 104.5652

18 104.317 104.517 104.5652

19 104.069 104.517 104.5652

20 104.751 104.517 104.5652

Average % of error 0.395 0.239


1956 R.K.Bhuyan et al./ Materials Today: Proceedings 4 (2017) 19471956

5. Conclusion

The present work is to compare the surface roughness such as Ra,Rq and Rz of Al-12% SiCp MMC during EDM
process by RSM and fuzzy logic technique. The RSM is an analytical method and fuzzy logic is a soft computing
method. Both the methods are simple and easy evaluation technique to predict the responses. The both methodology
give a close correlation between the predicted results and the experimental results but fuzzy logic technique give
less percentage of error in compare to RSM.Finally the significance of the each process parameter with the selected
response is analyzed by using ANOVA.

References

[1] J.S. Shaik, K.R.Babu, Int. J. of Emer. trends in Engg. and Develop., 5 (2012) 38-49.
[2] S.R. Das, D. Dhupal, A. Kumar, Int. J. of Pure and Appl. Research in Engg. andTech., 5 (2012) 25-36.
[3] R. K. Garg , K. K. Singh, A. S. Deva ,Vishal S. Sharma , K. Ojha, S. Singh,Int J Adv Manuf Technol 50(2010) 611 624.
[4] N.P Hung, L.J Yang, K.W Leong, J. of Mat. Processing Tech., 44 (1994) 229-236.
[5] D. M. Pargunde, Prof. G. N. Thokal, Prof. D. P. Tambuskar, Int. J. of Adv. Engg. Res. and Stud., 3 (2013) 22-25.
[6] Satyanarayen, D. Roystan,, M.Shreesaravanan, Balaguru, C.Devanathan, Int. J. of Adv. Res. Trend.in Engg. and Tech., (2015) 45-53.
[7] D. Puhan, S. S. Mahapatra, J. Sahu, L. Das, Measurement,46(2013) 35813592.
[8] M. K. Pradhan, C. K. Biswas,Int. J. Precision Technology, 2 (2010) 64-80.
[9] Rajesha S, C.S. Jawalkar, R. R. Mishra, A.K. Sharma, P. Kumar, Int. J. of Recent advan. in Mech. Engg., 3(2014) 53-62.
[10] M. Vishwakarma, V.K.Khare, V. Parashar, Int. J. of Engg. Research and Appl.,2(2012) 185-189.
[11] A.N.E. Khalid, A.N. Bakar, Sh.F. Ismail and M.S.N. Dout, Int. J. of systems Applications, Engg.& Develope, 5 (2011) 728-737.
[12] A. Kohli, A. Wadhwa, T. Virmani, U. Jain, World Academy of Science, Engg. and Tech., 72(2012) 1674-1679.
[13] P. S. Rao, K. E. Prasad, B. S. Reddy, International Journal of Research and Reviews in Applied Sciences, 9(2011)112-125.
[14] P. Shandilya, P.K.Jain, N.K. Jain, Procedia Engineering, 64(2013)767-774.
[15] D. Rodic, M. Gostimirovi, P. Kovac, M. Radovanovic,B. Savkovic, Int. J. of Recent adv. in Mech. Engg., 3 (2014) 69-77.
[16] B.S.Reddy, G.Padhanmava, K.V.K. Reddy, Asian journal of Scientific Res, 1(2008)256-264.
[17] M. S. Lou, J. C. Chen, C. M. Li,Surface Roughness Prediction Technique for CNC End-Milling.J. of Industrial Tech,15 (1999) 1-6.
[18] S. K. Majhi, T.K.Mishra, M.K.Pradhan, H. Soni , Int. J. of Current Engg. and Tech. , 4 (2014) 19-23.
[19] A.K. Sahoo, B. Sahoo,Int. J. of Industrial Engg. Computations, 2(2011) 469-478.
[20] R.K. Singh, D.K. Singh and V. Kumar, Int. Journal of Bio. Sci.& Tech. Research, 1 (2013) 37-41.
[21] J. Laxman and Dr.K.G. Raj, International J. of Adv. Mech. Engg., 4(2014) 473-480.
[22] P. Shandilya, P.K. Jain and N.K. Jain, Procedia Engineering, 38(2012) 23712377.
[23] A.K. Sahoo , K. Orra, B. C. Routra, Int. J. of Industrial Engg. Computations, 4(2013) 469-478.
[24] Vikas, A. K. Roy, K. Kumar, Procedia Mater. Sci.,6( 2014 )383390.
[25] S. Assarzadeh, M. Ghoreishi,Adv.Mater. Manuf. & Charac.3(2013)478-486.

S-ar putea să vă placă și